DE102017008158A1 - Numerical control - Google Patents

Abstract

A numerical controller, which controls a machine tool, acquires tool information including a shape of a tool, a machining condition in machining, and information related to a machining result of a workpiece after machining. A machine learning device performs machine learning on the tendency of the information regarding a machining result with respect to the tool information and the machining condition based on the tool information and the machining condition used as input data, and on the basis of the information relating to a machining result used as the teacher data to form a learning model. The machine learning device determines whether a machining result is good or not by using the learning model based on the tool information and the machining condition before the machine tool machines a workpiece.

Description

BACKGROUND OF THE INVENTION

1. Field of the invention

The present invention relates to a numerical controller, and more particularly relates to a numerical controller that learns a tool-condition causing machining error.

2. Description of the Related Art

• – Zu einem Zeitpunkt, zu dem die Zahl von Malen oder die Zeit, über die das Werkzeug verwendet wird, eine bestimmte Zahl überschreitet (Lebensdauerwert).
• – Zu einem Zeitpunkt, zu dem durch Messen der Werkzeuglänge festgestellt wird, dass eine Abnutzungsmenge eine bestimmte Zahl überschreitet.
• – Zu einem Zeitpunkt, zu dem ein Bediener durch optisches Prüfen des Werkzeugs oder eines bearbeiteten Werkstücks ermittelt, dass das Werkzeug das Lebensdauerende erreicht hat.
• – Zu einem Zeitpunkt, zu dem ein Brechen oder Absplittern durch Prüfen einer Last einer Spindel festgestellt wird.

In a normal tool used in the machining, a blade is worn, the cutting resistance increases with the lapse of the machining time and the tool breaks if necessary. Further, the machining accuracy decreases with the progress of wear, and thus, the required machining accuracy required for a workpiece can not be maintained. In order to avoid such problems with tools, it is mainly determined at the following times whether a tool has reached its end of service life, and a tool change is performed when it is determined that the tool has reached its end of life.

• - At a time when the number of times or the time over which the tool is used exceeds a certain number (lifetime value).
• At a time when it is determined by measuring the tool length that a wear amount exceeds a certain number.
• At a time when an operator determines that the tool has reached the end of life by visually inspecting the tool or a machined workpiece.
• At a time when breakage or chipping is detected by checking a load of a spindle.

As the prior art method for determining whether or not a tool has reached the end of life, for example, U.S. Pat Japanese Laid-Open Patent Application No. 2000-198047 a method in which the wear amount after machining is calculated based on a tool shape, a cutting condition, and an uncut amount of a workpiece to determine reaching an end of a tool life before machining.

However, the above-described method for determining a tool life has various problems. In the method of determining a tool life based on the number of times or the time that the tool was used, a lifetime value is generally set by more than a sufficient number of times of use or use time to avoid machining errors or breakage so that the tool is often still usable and costs arise. The method for determining tool life based on the tool length of a tool has a problem in that fractures of a chip can not be detected when a plurality of blades are present or a blade is disposed at the end. Further, in a tool whose service life is determined on the basis of wear in a portion other than the end portion, such as a tapping tool, tool life can not be determined by measuring the end portion. In the method for determining a tool life based on the optical inspection of a machined workpiece by an operator, an operator with technical knowledge must perform the operation, and therefore, this method is not suitable for unattended operation. In the method of determining a tool life by checking a spindle load, a slight chipping will hardly lead to changes in the spindle load, which causes difficulty in accurate detection.

Furthermore, in the method described above for determining a tool life as in Japanese Laid-Open Patent Application No. 2000-198047 described the accuracy in determining a tool life by taking into account a tool shape, a cutting condition and an uncut amount of a workpiece improved. The tool life can not be determined solely on the basis of the wear amount of a tool. The tool life is also determined by a material of the tool, a position of wear, a wear condition (such as chipping) and the like. Ä. affected. Thus, the method of the Japanese Laid-Open Patent Application No. 2000-198047 the problem is that the method is not designed for different situations.

SUMMARY OF THE INVENTION

It is an object of the present invention to provide a numerical controller that enables highly accurate tool life determination.

In the present invention, the above-described problems are solved by providing numerical control having a function of learning a relationship between a machining condition and a machining result by supervised learning and improving the input data based on the determined learning result. The numerical controller according to the present invention may include a While maintaining some level of quality by improving input data based on learning outcomes.

A numerical controller according to the present invention controls a machine tool that processes a workpiece with a tool, and includes: a shape data acquiring unit that acquires tool information including a shape of the tool; a machining condition detecting unit that includes a machining condition in the machining of the workpiece; a machining result detecting unit that acquires information regarding a machining result of the workpiece after machining the workpiece; a machine learning device that performs machine learning on the tendency of the information related to a machining result with respect to the tool information and the machining condition on the basis of the tool information used as the input data and the machining condition, and on the basis of the information relating to a machining result used as the teacher data; to form a learning model. The machine learning device is configured to determine whether a machining result is good or not by using the formed learning model based on the tool information acquired by the shape data acquiring unit and the machining condition detected by the machining condition acquiring unit before the machine tool processes the workpiece.

The machine learning device may form a learning model through a multi-layered neural network.

Thus, the numerical controller according to the present invention has the above-described configuration. Accordingly, a reduction in tooling costs due to the improvement in tool life determination is to be expected, and unattended, continuous operation can be promoted.

BRIEF DESCRIPTION OF THE DRAWINGS

1 Fig. 10 shows an overview of machine learning and the application of a learning result by a machine learning device incorporated in a numerical controller in the present invention.

2 FIG. 16 shows a configuration example (learning phase) of the machine learning device incorporated in the numerical controller according to the present invention. FIG.

3 shows the numbering of images, sectional views and materials of tools as well as the numbering of materials of a workpiece.

4 FIG. 16 shows a configuration example (operation phase) of the machine learning apparatus incorporated in the numerical controller according to the present invention. FIG.

5 FIG. 12 is a diagram (FIG. 1) which is a schematic functional block diagram of the numerical control in the learning phase according to an embodiment of the present invention.

6 10 is a diagram (2) which is a schematic functional block diagram of the numerical control in the learning phase according to the embodiment of the present invention.

7 FIG. 3 is a flowchart showing an operation of the in 5 and 6 represented numerical control.

8th FIG. 12 is a schematic functional block diagram of the numerical control in the operation phase according to the embodiment of the present invention. FIG.

9 FIG. 3 is a flowchart showing an operation of the in 8th represented numerical control.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

1 Fig. 10 shows an overview of machine learning and the application of a learning result by a machine learning device incorporated in a numerical controller in the present invention. The machine learning device integrated in the numerical controller according to the present invention operates in two phases, which are a learning phase and an operating phase.

In the learning phase, the machine learning apparatus incorporated in the numerical controller according to the present invention uses as an input data processing conditions such as tool information, a material of a workpiece, a cutting speed, a rotational speed of a spindle, an amount of the cutting depth and an intersection, and a machining result such as a dimensional error of one the basis of the machining conditions machined workpiece for learning a tendency of the machining result in relation to the machining conditions.

In the operating phase, the numerical control according to the present invention provides Processing conditions as input data for the machine learning device. The machine learning device outputs a predicted value of a machining result with respect to the input machining conditions with respect to the tendency of the machining result with respect to the machining conditions learned in the learning phase. Subsequently, the numerical controller refers to the predicted value of a machining result output from the machine learning device to determine whether an error will occur in a subsequent machining operation.

In the numerical control according to the present invention, the learning phase and the operation phase of the machine learning device can be executed at completely different timings or in parallel. For example, after the machine learning apparatus is allowed to sufficiently perform learning in the learning phase, the operation can be performed by using only the learning results without performing the new learning, that is, by repeatedly using the same data in the current machining by a machine tool. Alternatively, the operation may be performed so that more accurate judgment values are searched (learned) during the processing.

Further, the machine learning device may be incorporated in the numerical control of a machine tool or may be incorporated into a PC connected to the numerical controller.

2 and 3 show a configuration example of the machine learning apparatus incorporated in the numerical controller according to the present invention.

2 and 3 show the configuration in which a multilayer neural network is used as the machine learning device. The multilayer neural network consists of an arithmetic device, a memory and the like. Ä., Which run a designed as an imitation of a model of a neuron neural network.

As in 2 When the multilayer neural network is used as the machine learning apparatus of the numerical controller according to the present invention, the numerical controller of the present invention uses machining conditions in the machine learning apparatus as input data (X1 to X6 in FIG 2 ) in the learning phase before processing in each step (or at a time when a tool or cutting condition is changed, or at any time). Examples of the machining condition include tool information (image, sectional view, material), a material of a numbered workpiece, a cutting speed (combined speed), a rotational speed of a spindle, an amount of cutting depth, and an intersection. Here, for an image, a sectional view, or a material of a tool and a material of a workpiece, a plurality of typical image data and material data may be previously numbered and stored, as in FIG 3 and numbers may be used as input data with respect to the machine learning device. Thus, the number of input data can be reduced, and the structure of the multilayer neural network can be simplified.

Subsequently, the numerical controller according to the present invention detects output data W output from the machine learning device when the processing conditions serving as input data are input to the multilayer neural network as a predicted processing result. Further, after a workpiece is machined on the basis of the machining conditions input to the multilayered neural network, the numerical controller detects the dimension of the machined workpiece, the number and size of the cutter marks, and the like. Ä. as a machining result, and calculates an error D between the detected machining result and a design as teacher data. The numerical control can automatically determine a measure of the machined workpiece, the number and size of the cutting tool marks and the like. Ä. for example, by controlling a measuring instrument or an operator can perform the measurement of the machined workpiece and enter the measured result in the numerical controller.

Then, the numerical controller according to the present invention enables the machine learning device to perform the machine learning with the machining results with respect to the machining conditions based on the machining conditions serving as input data, the predicted machining result output from the multilayer neural network, and the error D (teacher data) the dimension of the machined workpiece serving as the detected machining result and the design for forming a learning model. Since the learning method (backpropagation or the like) is already well known as a prior art method with respect to a neural network, a detailed description thereof is omitted in the present specification. The learning is performed until a distance between output data and a machining result is set within a predetermined specific value Has. The distance is calculated by using a loss function. As an example of the loss function, a quadratic error (L2 loss) may be used as expressed in the following formula (1), for example, when focusing on an error between a dimension of a machined workpiece and a design value. Further, when focusing on the presence of a cutter mark, a logical function that merely expresses whether a cutter mark is present or not can be used. Furthermore, any function can be used as long as the function can sufficiently express a loss. L (Y (Xn; W), D) = (Y (Xn; W) - D) ² (1)

As in 4 When the multilayer neural network is used as the machine learning apparatus of the numerical controller according to the present invention, the numerical control indicates machining conditions in the machine learning apparatus as input data (X1 to X6 in FIG 2 1) before machining in each step (at a time when a tool or a cutting condition is changed, or at any time) in the operation phase. As elements of the machining conditions input to the machining apparatus in the operation phase, elements similar to those of the machining conditions used in the learning phase are used.

Then, the numerical controller detects the output data W output from the machine learning device when the machining conditions serving as input data are input to the multilayer neural network as a predicted machining result (an error of a predicted dimension of a machined workpiece with respect to a design). the presence of a cutter mark, etc.).

Then, if it is determined that a result of the processing is bad, when the predicted processing result serving as the output data is compared with a predetermined threshold value (an error of a predicted amount of a machined workpiece with respect to a design is equal to a predetermined specific value or greater as such, the number of cutter marks is equal to or larger than a predetermined number, the size of a cutter mark is equal to or larger than a predetermined size, etc.), the numerical controller according to the present invention that the machining result is a machining detection pattern, and prompts an operator to perform a tool change.

5 and 6 FIG. 12 are functional block diagrams showing essential parts of a numerical controller according to an embodiment of the present invention in the learning phase. Further shows 7 a flowchart illustrating a sequence of operation of a numerical control 1 in the learning phase.

The numerical control 1 According to the present embodiment, a machine learning device includes 10 a form data acquiring unit 11 , a machining condition detecting unit 12 and a machining result detection unit 13 , Furthermore, the machine learning device includes 10 an input condition observer 110 , an input condition storage unit 120 , a unit for calculating a predicted machining result 130 , a unit for storing a predicted processing result 140 , a processing result observation unit 150 , a processing result storage unit 160 and a comparison unit 170 ,

The numerical control 1 Switches to the learning phase and begins learning when the time to complete the learning of the machine learning device 10 has come (step SA01). Examples of the timing of switching to the learning phase include a timing at which a tool is changed to use, a timing at which a machining condition (a machining content, a cutting speed, a rotational speed, or the like) is changed, and a user fixed arbitrary time.

The shape data acquisition unit 11 and the machining condition detecting unit 12 each guide the detection of tool information (shape information (an image, a sectional view), a material of a tool, etc.) and the detection of machining conditions (a material of a workpiece, machining contents (flat surface / side surface, etc.), a rotational speed, a cutting speed, an intersection) by (steps SA02 and SA03). Tool information and machining conditions may be based on setpoints or a numerical control program 1 detected or entered by an operator. Here, for shape information and a material of a tool and a material of a workpiece, numerical values as in 3 shown captured. In particular, for the shape information of a tool, a tool attached to a machine tool may be subjected to imaging by an imaging device, not shown, and image processing may be performed to acquire an image representing a shape of the tool to be used. Then that can captured image with typical tool image data as in 3 can be adjusted to use a number of image data to represent the next shape as the tool information.

The input condition observer 110 detects the from the shape data acquisition unit 11 detected tool information and that of the machining condition detection unit 12 detected processing conditions as input conditions of the machine learning device 10 , The input condition observer 110 observes detection states of each of the shape data detection unit 11 and the machining condition detecting unit 12 collected data. When the shape data acquiring unit 11 and / or the machining condition detecting unit 12 Capture data captures the input condition observer 110 the data as input data.

The input condition storage unit 120 saves from the input condition observer 110 detected input conditions on a memory, which is not shown. The input condition storage unit 120 can only use the input conditions provided by the input condition observer 110 be monitored and recorded by overwriting and storing input conditions stored in the memory. Thus, only input conditions that change can be updated.

The unit for calculating a predicted machining result 130 calculates a predicted machining result on the basis of the input condition storage unit 120 stored input conditions and one of the unit for storing a predicted processing result ( 140 ) (learning model) (step SA04). The unit for storing a predicted machining result 140 For example, it consists of a multilayered neural network as previously described.

After the numerical control 1 a workpiece is being machined while using a machine tool 2 on the basis of the processing condition acquiring unit 12 and a machining program, which is not shown, controls (step SA05), detects the machining result detection unit 13 a machining result of the machined workpiece (step SA06). A machining result to be detected includes an error between a measurement of the machined workpiece and a design dimension, the presence of a cutting tool mark, and the like. Ä.

The processing result observation unit 150 detects the machining result obtained by the machining result detection unit 13 in the machine learning device 10 , The processing result observation unit 150 observes a detection state of all of the machining result detection unit 13 collected data. When a machining result is detected, the machining result observing unit detects 150 the data in the machine learning device 10 ,

The processing result storage unit 160 stores this from the machining result observing unit 150 recorded processing result on a memory, which is not shown.

The comparison unit 170 compares this with the unit for calculating the predicted machining result 130 calculated predicted machining result with that of the machining result observing unit 150 captured (and from the processing result storage unit 160 stored) actual machining result and calculates a difference between the predicted machining result and the actual machining result by using a loss function to compare the difference with a predetermined threshold (step SA07). If the difference between the predicted machining result and the actual machining result is within the predetermined threshold, no learning is performed. When the difference is equal to or greater than the threshold, learning is performed on the basis of the predicted processing result and the actual processing result, and the determined learning result is stored in the unit for storing a predicted processing result 140 stored (steps SA08 and SA09).

8th FIG. 11 is a functional block diagram showing essential parts when the in 5 and 6 described numerical control 1 can work in the operating phase. Further shows 9 a flow chart illustrating a sequence of operation of the numerical control 1 in the operating phase. In the operating phase includes the numerical control 1 a tool change instruction unit 14 and the machine learning device 10 further comprises a tool change detection unit 180 ,

In the operating phase, the numerical control leads 1 According to the present embodiment, an operation of a preprocessing test with the machine learning device 10 through, when the numerical control 1 in a state for performing the pre-processing check (step SB01). Examples of the state for performing the pre-processing check include a time at which a tool is changed to use, a timing at which a machining condition (a machining content, a cutting speed, a rotational speed, or the like) is changed, and an arbitrary timing set by a user.

The shape data acquisition unit 11 and the machining condition detecting unit 12 each guide the detection of tool information (shape information (an image, a sectional view), a material of a tool, etc.) and the detection of machining conditions (a material of a workpiece, machining contents (flat surface / side surface, etc.), a rotational speed, a cutting speed, an intersection) by (steps SB02 and SB03).

The input condition observer 110 detects the from the shape data acquisition unit 11 detected tool information and that of the machining condition detection unit 12 detected processing conditions as input conditions of the machine learning device 10 ,

The input condition storage unit 120 saves from the input condition observer 110 detected input conditions on a memory, which is not shown.

The unit for calculating a predicted machining result 130 calculates a predicted machining result on the basis of the input condition storage unit 120 stored input conditions and one of the unit for storing a predicted processing result 140 stored learning content (step SB04).

The above-described operations by the shape data acquiring unit 11 , the machining condition detecting unit 12 , the input condition observer 110 , the input condition storage unit 120 , the unit for calculating a predicted machining result 130 and the unit for storing a predicted processing result 140 are the same as those performed by the respective functional means in the learning phase.

The tool change detection unit 180 performs the tool change determination based on that of the predicted machining result calculating unit 130 calculated predicted processing result by. The tool change detection unit 180 refers to that of the unit for calculating a predicted machining result 130 outputted predicted machining result (an error between a dimension of a machined workpiece and a design value, the presence of a cutter mark, etc.), and when the result of machining is poor (the error is equal to or greater than a predetermined predetermined threshold, the number of cutter marks is equal to or greater than a predetermined number, the size of a cutting edge mark is equal to or larger than a given size, etc.), the tool change determining unit determines 180 in that the tool needs to be changed (step S605).

On the basis of the determination result of the tool change determination unit 180 if it is determined that the tool needs to be changed, the tool change instruction unit 14 replacing the tool (step SB06). The tool change instruction unit 14 may be a tool change instruction with respect to the machine tool 2 to output tools for automatically changing tools, or to perform a tool change request display on a display device of a control panel, not shown, to allow an operator to look at the tool change display.

The embodiment of the present invention has been described so far. However, the present invention is not limited to the above-described example of the embodiment, and can be carried out by making corresponding changes.

For example, the example is using the multilayer neural network as the machine learning device 10 described in the previous embodiment. However, various methods such as the least squares method, stepping method, SVM, and decision tree learning are also well-known supervised learning algorithms, and each of these methods can be used as the machine learning device 10 of the present invention.

Furthermore, as with respect to 3 described pictures for representing tool shapes in advance numbered and numerical control 1 stored according to the present invention. This numbering can be done by manual operations; automation may be easier so that clustering is performed by unmonitored learning with respect to a plurality of tool images, representative images are determined based on the result of the clustering, and the numbering is performed with respect to each of the representative images.

Further, the operation of the numerical control 1 in the learning phase and the operation of numerical control 1 in the operating phase individually in relation to 5 to 9 described; however, the learning phase and the operating phase can also be carried out in parallel. In such a case, the pre-processing check is performed in the operating phase, and processing result information after completion of the processing is detected at the same time with the machine learning device whose learning has progressed to a certain level. In such a state that a difference between a predicted machining result and an actual machining result is equal to or greater than a predetermined threshold, learning with respect to the predicted machining result storing unit may be made 140 respectively.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• JP 2000-198047 [0003, 0005, 0005]

Claims (2)

A numerical controller that controls a machine tool processing a workpiece with a tool, wherein the numerical controller comprises: a shape data acquisition unit that acquires tool information including a shape of the tool; a machining condition detecting unit that includes a machining condition in the machining of the workpiece; a machining result detecting unit that acquires information regarding a machining result of the workpiece after machining the workpiece; and a machine learning device that performs machine learning on the tendency of the information related to a machining result with respect to the tool information and the machining condition on the basis of the tool information used as the input data and the machining condition, and on the basis of the information relating to a machining result used as the teacher data; to form a learning model, where the machine learning device for determining whether a machining result is good or not is formed by using the formed learning model based on the tool information acquired by the shape data acquisition unit and the machining condition detected by the machining condition detection unit before the machine tool processes the workpiece. The numerical controller of claim 1, wherein the machine learning device forms a learning model through a multilayer neural network.

Images